1.5 Million Missing Numbers: Overcoming Employment Suppression in County Business Patterns Data

Abstract

Missing data frustrate research and limit our understanding of regional economies. County Business Patterns annually provides employment data for all U.S. counties and states at the most detailed industrial level, but two out of every three employment statistics are missing. In rural areas, this percentage is higher still. To protect the rights of employers to confidentiality, the U.S. Census Bureau has not disclosed the number of employees in 1.5 million cases in the 2002 data. Instead, it offers a suppression flag that represents an employment range. This article presents a two-stage method for replacing all the flags with employment estimates. Taking advantage of the hierarchical nature of the data both by industry and geography, the first stage identifies the smallest possible range for each suppressed number. Ensuring that employment adds up correctly up and down the industrial and geographical hierarchies, the second stage iteratively adjusts all the estimates until millions of constraints are met. The procedure simultaneously considers all industries in all counties, states, and the nation to produce a complete data set, which is available to the research community on the Internet.